[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] Varying the takepoint for money play
From: |
Christian Anthon |
Subject: |
Re: [Bug-gnubg] Varying the takepoint for money play |
Date: |
Wed, 14 Feb 2007 10:46:40 +0100 |
we had a thread recently about 0 ply doubling early. You might want to
go back and read that. Anyway, the neural net is strictly cubeless, so
what 0ply does is estimate the cubefull equity by using a cube
effectiveness factor:
/* parameters for EvalEfficiency */
float rOSCubeX = 0.6f;
float rRaceFactorX = 0.00125f;
float rRaceCoefficientX = 0.55f;
float rRaceMax = 0.7f;
float rRaceMin = 0.6f;
float rCrashedX = 0.68f;
float rContactX = 0.68f;
defined in eval.c. So you may either change these numbers or change
the live take point of 1.0 (which I haven't checked how to do).
It is probably easier to set two instances of gnubg up to play against
each other through something like dueller than it is to program one
instance to use different parameters for the two players.
Christian.